The Best American Science Writing 2006

Read an Excerpt

The Best American Science Writing 2006

HarperCollins Publishers, Inc.

Your Move

Tom Mueller

from The New Yorker

When the IBM supercomputer Deep Blue defeated chess champion Garry Kasparov in 1997, it was heralded as a major turning point in the continuing struggle between man and machine. Adapting Deep Blue's approach to less powerful PCs, programmers are making up for lack of number-crunching ability with artfulness. As Tom Mueller has discovered, the results have been unexpected, with computers developing strategies grand masters have never thought of.

Chrilly Donninger prefers to watch from a distance when Hydra, his computer chess program, competes, because he is camera-shy, but also because he rarely understands what Hydra is doing, and the uncertainty makes him nervous. During Hydra's match against the world's seventh-ranked player, Michael Adams, in London last June, Donninger sat with three grand masters at the back of a darkened auditorium, watching a video projection of the competition on the wall behind Adams. Most of the time, Donninger, a forty-nine-year-old Austrian, had little to worry about; Hydra won the match five games to none, with one draw. But in the second game, which ended in the draw, the program made an error that briefly gave its human opponent an advantage.

The game was played at a spotlit table on a low podium. Adams sat in theclassic chess player's pose--his elbows resting on the table, his chin cupped in his palms--reaching out now and then with his right hand to move a piece on a large wooden chessboard. Across from him was Hydra--a laptop linked by Internet connection to a thirty-two-processor Linux cluster in Abu Dhabi--and Hydra's human operator, who entered Adams's moves into the computer and recorded the program's replies on the board. On the laptop's screen was a virtual chessboard showing the current position in the game, as well as a pane of swiftly scrolling numbers representing a fraction of the thousands of lines of play that Hydra was analyzing, and a row of colored bars that grew or shrank with each move, according to the program's assessment of who was winning--green bars meant an advantage for white, red bars for black.

For much of the match, the bars showed Hydra comfortably in the lead. When Adams made a mistake, they spiked dramatically, but mostly they grew in small increments, recording the tiny advantages that the program was steadily accumulating. Many of these were so subtle that Donninger and the grand masters failed to grasp the logic of Hydra's moves until long after they had been made. But about twenty minutes into the second game, when Hydra advanced its central e-pawn to the fifth rank, there was a small commotion in the group. Yasser Seirawan, an American player formerly ranked in the top ten, who had coached Adams for the match, gave a thumbs-up sign. Christopher Lutz, a German grand master who is Hydra's main chess adviser, groaned. Only Donninger, who programs chess far better than he plays it, was baffled. He turned to Lutz in alarm.

"What was that? What did you see?"

"Now our pawn structure has become inflexible," Lutz replied. "Do we have anything in the program for flexibility?"

"What do you mean by 'flexibility'?"

Lutz frowned. He sensed that Hydra had hemmed itself in, giving Adams the upper hand. Bishop to b7 was the correct move, Lutz believed--the most natural way for Hydra to preserve its attacking chances and its room to maneuver. But explaining his nebulous insights to a lesser player like Donninger was a challenge.

"This position lacks flexibility," he repeated, shaking his head.

"When you can define 'flexibility' in twelve bits, it'll go in Hydra," Donninger told him, twelve bits being the size of the program's data tables.

Adams locked up Hydra's center with his next move and managed, several hours later, to eke out a draw. "Hydra didn't play badly, but 'not bad' isn't good enough against a leading grand master," Donninger said after the game. His program is widely considered to be the world's strongest chess player, human or digital, but it still has room for improvement.

Lean and restless, with a scraggly beard and a large Roman nose, Donninger says that he approaches programming less like a scientist than like a craftsman--he compares himself to a Madonnenschnitzer, one of the painstaking Baroque and rococo wood-carvers whose Madonna sculptures adorn the churches near Altmelon, the village in northern Austria where he lives and works. He speaks German with a thick Austrian brogue and frequently uses expressions like "Das ist mir Wurscht!"--"That's all sausage to me!" For the past two years, he has led the Hydra project, a multinational team of computer and chess experts, which is funded by the Pal Group, a company based in the United Arab Emirates which makes computer systems, desalinization plants, and cyber cafes. Pal's owner, Sheikh Tahnoon bin Zayed al-Nahyan, is a member of the country's royal family and a passionate chess player; he hired Donninger with the goal of creating the world's best chess program. Pal is also using the same kind of hardware that runs Hydra for fingerprint-matching and DNA-analysis applications, which, like computer chess, require high-speed calculations. The program's main hardware resides in an air-conditioned room in Abu Dhabi, and Donninger is frequently unable to access it, because the sheikh and Hydra, playing under the name zor_champ, are on the Internet, taking on all comers.

As a child, Donninger was so attached to puzzles that his mother worried that he was disturbed. At the age of four, he spent months building houses out of four colors of Lego bricks, in which no bricks of the same color ever touched; two decades later, when he was an undergraduate at the University of Vienna, he learned that this was a famous conundrum in topology--the Four-Color Problem. After completing a doctorate in statistics, he worked as a programmer for Siemens, where he earned a reputation as a bug fixer, the computer equivalent of a puzzler. In 1989, he was transferred . . .

---

Excerpted from The Best American Science Writing 2006 by Atul Gawande Copyright © 2006 by Atul Gawande. Excerpted by permission.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.

---